A New Population Initialization Method Based on Space Transformation Search

Abstract

Population initialization, as an important step in population-based stochastic algorithm, can affect the convergence speed and the quality of solutions. Generally, random initialization is used to generate initial population when lacking priori information. This paper presents a new initialization method by applying space transformation search (STS) strategy to generate initial population. Experimental results on 8 well-known benchmark problems show that the population initialization based on STS outperforms traditional random initialization and opposition-based population initialization.